When measuring knee kinematics with skin-mounted markers, soft tissue and structures surrounding the knee hide the actual underlying segment kinematics. Soft-tissue artefacts can be reduced when plate-mounted markers or marker trees are used instead of individual unconstrained mounted markers. The purpose of this study was to accurately quantify the soft-tissue artefacts and to compare two marker cluster fixation methods by using fluoroscopy of knee motion after total knee arthroplasty during a step-up task. Ten subjects participated 6 months after their total knee arthroplasty. The patients were randomised into (1) a plate-mounted marker group and (2) a strap-mounted marker group. Fluoroscopic data were collected during a step-up motion. A three-dimensional model fitting technique was used to reconstruct the in vivo 3-D positions of the markers and the implants representing the bones. The measurement errors associated with the thigh were generally larger (maximum translational error: 17 mm; maximum rotational error 12°) than the measurement errors for the lower leg (maximum translational error: 11 mm; maximum rotational error 10°). The strap-mounted group showed significant more translational errors than the plate-mounted group for both the shank (respectively, 3±2.2 and 0±2.0 mm, p=0.025) and the thigh (2±2.0 and 0±5.9 mm, p=0.031). The qualitative conclusions based on interpretation of the calculated estimates of effects within the longitudinal mixed-effects modelling evaluation of the data for the two groups (separately) were effectively identical. The soft-tissue artefacts across knee flexion angle could not be distinguished from zero for both groups. For all cases, recorded soft-tissue artefacts were less variable within subjects than between subjects. The large soft-tissue artefacts, when using clustered skin markers, irrespective of the fixation method, question the usefulness of parameters found with external movement registration and clinical interpretation of stair data in small patient groups. © 2007 Elsevier Ltd. All rights reserved.